The membrane-proximal external region of human immunodeficiency virus (HIV-1) envelope glycoprotein trimers in A18-lipid nanodiscs

Purification and characterization of the cleaved full-length Tri FPPR Env

The full-length Tri FPPR Env, which contains changes that favor the PTC of membrane Env¹⁸, was expressed in A549 human lung epithelial cells (Supplementary Fig. 1a, b). Membranes prepared from these cells were used for Env purification. BMS-806, which slows Env transitions from the PTC^5,11, was added to the membranes and was present throughout the purification of the Tri FPPR Env. The Tri FPPR Env was extracted directly from the membranes with Amphipol A18^43,44, purified using the carboxy-terminal six-histidine (His₆) affinity tag, and subjected to counterselection with the 19b and F240 poorly neutralizing antibodies to remove uncleaved Envs and gp41 stumps that have shed gp120. Approximately 96% of the purified Tri FPPR Env in the preparation was cleaved (Supplementary Fig. 1c). Tri FPPR Env trimers were evident on a Blue Native polyacrylamide gel. The purified cleaved Tri FPPR Env was recognized by a panel of bNAbs and not by a panel of pNAbs; this pattern of antibody recognition was similar to that of the Tri FPPR Env on the surface of A549 cells⁴³.

Two conformers of asymmetric Tri FPPR Env structures determined by cryo-EM

We used single-particle cryo-EM to analyze the structure of the Tri FPPR Env in A18-lipid nanodiscs. Extensive 2D and 3D analysis without imposing symmetry constraints yielded two 3D maps notable for their recognizable MPER density (Supplementary Fig. 2 and see below). These maps are designated Tri FPPR.1 and Tri FPPR.2, with average resolutions of 3.4 Å and 3.5 Å, respectively (Fig. 1, Table 1, Supplementary Fig. 3). The quality of the Tri FPPR.1 and Tri FPPR.2 maps allowed atomic modeling and refinement with accuracy to the level of the Cα backbone trace and some large side chains (Supplementary Fig. 4a). With the exception of certain glycans in the gp120 variable regions (modifying Asn 136, 141 and 142 in V1; Asn 406 and 411 in V4; and Asn 463 in V5), the density associated with the Env glycans was apparent, allowing modeling of the peptide-proximal carbohydrate residues (For example, see Supplementary Fig. 4b). Ectodomain residues 31-662 of the Tri FPPR.1 and Tri FPPR.2 trimers share an overall topology with each other and with those of existing soluble and membrane Env trimer structures^{25,38,39,40,41,42,45,46,47,48}. In all these structures, the gp120 subunits project from the gp41 subunits, with the C-terminal segments of the heptad repeat 1 (HR1_C) regions of gp41 forming a three-helix bundle. The gp41 MPER (residues 663-683), transmembrane (TM) region (residues 684-705) and cytoplasmic tail (CT) are not present in sgp140 SOSIP.664 trimers³¹ and are not well resolved in available structures of detergent-solubilized Env trimers, including those reconstituted into lipid-bilayer nanodiscs^{25,38,39,40,41,42}. The Tri FPPR.1 and Tri FPPR.2 maps both exhibited significant density corresponding to the MPER and N-terminal TM, with average resolutions of 8 Å and 6-8 Å, respectively. The CT and C-terminal portions of TM were not resolved in either the Tri FPPR.1 or Tri FPPR.2 maps.

a. The two schematic diagrams on the left indicate the nomenclature of the gp120 chains (chains A, C and E) and gp41 chains (chains B, D and F). Except where noted, the color scheme for the Env protomers and subunits will be used throughout the manuscript. The schematic diagram on the left is a view down the Env trimer axis, from the perspective of the membrane of the virus or expressing cell. The opening angle between Protomer 1 and Protomer 3 is less than 120 degrees in these asymmetric trimers. The schematic diagram on the right is a side view, with the gp41 subunits at the top and the gp120 subunits at the bottom of the images. On the right, two 5 Å-filtered Tri FPPR.2 density maps are shown, from the same perspectives used for the schematic diagrams. The α9 helix and MPER from Protomer 1 (red ribbons) interacts with the FPPR-HR1_N structures (blue ribbons) of the clockwise protomer (Protomer 3), when viewed from the perspective of the viral/cell membrane. b The Tri FPPR.1 and Tri FPPR.2 density maps are colored and shown from the same perspectives as in (a). c The Tri FPPR.1 and Tri FPPR.2 models are shown, fitted into the respective maps that were filtered to 5-Å resolution.

The Tri FPPR.1 or Tri FPPR.2 Env trimers lack C3 symmetry due to differential rotation of the protomers with respect to the trimer axis. In these asymmetric trimers, the adjacent gp120 subunits exhibit the following geometric relationship: one opening angle is less than 120° and two opening angles are greater than 120° (Fig. 1a). The same pattern of asymmetry has been observed in uncleaved HIV-1_JR-FL Env trimers and cleaved HIV-1_AD8 and AE2 Env trimers purified from solubilized membranes^25,41 (Supplementary Fig. 5) despite differences in Env sequence, cleavage and preparation (Supplementary Fig. 1a).

Beginning with a C3-symmetric Env trimer, the asymmetric AD8, AE2, Tri FPPR.1 and Tri FPPR.2 Env trimers can be derived by rotational shifts of Protomers 2 and 3 about the trimer axis (Supplementary Fig. 6a, Table 2). In all these cases, Protomers 2 and 3 shift in a counterclockwise direction when the Env trimer is viewed from the perspective of the viral membrane. The shift in Protomer 3 exceeds that of Protomer 2, resulting in asymmetric Env structures in which the opening angle between Protomers 1 and 3 is <120° and the other two opening angles are >120°. This pattern of asymmetric Env opening has been suggested to assist the binding of two CD4 molecules, which are required for HIV-1 entry^41,49. The shift in the gp41 subunit of Protomers 2 and 3 exceeds that of the gp120 subunit. This observation is consistent with changes in gp41 driving the shift, with the gp120 subunits following suit. In this manuscript, we adhere to the previously published designations of the asymmetric Env protomers and chains⁴¹ (Fig. 1a); in this scheme, the smaller opening angle is between Protomers 1 and 3.

We note that the observed trimer asymmetry is not explained by non-uniform occupancy of BMS-806 in the previous study⁴¹ or in the Tri FPPR Envs (Supplementary Fig. 4c).

Structures of the Tri FPPR.1 and Tri FPPR.2 Envs

The structures of the Tri FPPR.1 and Tri FPPR.2 Env protomers were compared (Fig. 2). The gp120 subunits exhibited similar folds, resembling those observed in previously reported Env trimer structures^{25,38,39,40,41,42,45,46,47,48}. When the gp120 subunits were aligned, the interprotomer Cα RMSD was 0.63-0.75 Å for Tri FPPR.1 and 0.71-0.88 Å for Tri FPPR.2. Conformational differences among the three protomers were evident in the gp41 fusion peptide, fusion peptide-proximal region (FPPR), N-terminal heptad repeat (HR1_N) and α9 helices. As has been previously seen for the HIV-1_AD8 and AE2 Env trimers⁴¹, the configuration of these gp41 elements differed among the protomers of the asymmetric Tri FPPR trimers. Notably, helical HR1_N regions are present in Protomers 1 and 2, whereas HR1_N is a loop in Protomer 3. Interprotomer contacts between HR1_N and α9 apparently influence HR1_N conformation. Differences in HR1_N helicity correlate with the tilt angles of the α9 helices on the adjacent protomer and with the degree of order in the MPERs extending from those α9 helices (Fig. 2 and Table 2). Thus, when Env is viewed down the trimer axis from the perspective of the membrane (as in the leftmost image in Fig. 1a), the α9-MPER conformation on each gp41 protomer is related to the FPPR-HR1_N structures on the clockwise protomer (Fig. 1a and Supplementary Fig. 6b).

a–c Molecular structures of the three protomers of the Tri FPPR.1 and Tri FPPR.2 Env models were aligned using the gp120 subunits. Conformational differences among the protomers were evident in the gp41 fusion peptide (magenta), the fusion peptide-proximal region (FPPR) and N-terminal heptad repeat (HR1_N) (blue), the α9 helices (orange) and the MPERs (salmon). In contrast, the gp41 HR1_C helices (red) of the three protomers exhibited similar conformations. a The three aligned protomers of the Tri FPPR.1 model are shown in the left panel, and the three aligned protomers of the Tri FPPR.2 model are shown in the right panel. The insets show the gp41 subunits from a different perspective. b. In the left panel, Protomer 2 (darker shading) and Protomer 1 (lighter shading) of the Tri FPPR.1 model are compared. In the right panel, Protomer 3 (darker shading) and Protomer 1 (lighter shading) of the Tri FPPR.1 model are compared. c Protomers 1, 2 and 3 of the Tri FPPR.2 model (darker shading) are compared with the corresponding protomers of the Tri FPPR.1 model (lighter shading).

Remarkable differences among the Tri FPPR Env protomers were observed in the map densities associated with the gp41 MPERs. The gp41 MPER extends from the carboxyl terminus of the α9 helix (~Glu 662-Asp 664) to the transmembrane (TM) region, which begins at Ile 684. For the Tri FPPR.2 Env, the entire MPER as well as six TM residues of Protomer 1 (chain B) could be modeled (Fig. 3a). The MPER of Tri FPPR.2 chain B is composed of two helices, MPER_N and MPER_C, separated by a short turn and flanked by α9 and TM (Fig. 3 and Supplementary Fig. 7a). Immediately following the α9 helix of chain B, a sharp 71-degree turn at residues 664 and 665 directs the MPER_N helix towards the α9 helix of the adjacent protomer [Protomer 3 (chain F)]. Sandwiched between the chain B MPER_N helix and chain F α9 helix is the C-terminus of chain E gp120. The MPER_N helix breaks at Thr 676-Asn 677, introducing a slight turn that directs the MPER_C helix to the membrane. The initial few TM residues (Ile 684-Val 689) appear to retain a helical conformation. In contrast to the map density associated with the Tri FPPR.2 Protomer 1 (chain B), the Tri FPPR.2 map density corresponding to the MPERs in Protomer 2 (chain D) and Protomer 3 (chain F) was very weak; therefore, the chain D and F MPERs were not included in the Tri FPPR.2 Env model.

a Ribbon models of the membrane-proximal structures of Tri FPPR.1 Protomer 2 and Tri FPPR.2 Protomer 1 are shown with the corresponding maps filtered to 5 Å. Note the proximity of the gp120 C-terminus to the gp41 α9-MPER_N helices of the same protomer in each structure. b The α9 (orange), MPER (salmon) and TM (blue) of Tri FPPR.1 (left panel) and Tri FPPR.2 (right panel) are shown from the same perspective, with the turn angles between α9 and MPER_N depicted. In the left panel, the α9 and MPER of Tri FPPR.1 chains D and B are compared. In the right panel, the α9 and MPER of Tri FPPR.2 chain B are shown. c Ribbon models of the Tri FPPR.1 and Tri FPPR.2 Env trimers are shown on the left with the subunits colored as in Fig. 1 and the MPERs at the top. In the enlarged figures on the right, the membrane-proximal base of each trimer is shown from the perspective of the membrane, looking down the trimer axis. The smallest opening angle between the gp120 subunits is at the bottom of the images. In each case, the well-ordered MPERs project from the α9 helix of their own chain towards the α9 helix of the adjacent protomer. The MPER_C extends to the apex of the adjacent protomers, where the TM enters the membrane. In the Tri FPPR.1 structure, one face of the membrane-proximal base of Env is highlighted, illustrating how the gp120 C-terminus is sandwiched between the gp41 α9 helix of the same protomer and the MPER_N of the adjacent protomer. This interlocking arrangement creates opportunities for stabilizing bonds among the Env protomers and subunits.

For Tri FPPR.1, the entire MPER of Protomer 2 (chain D) and the MPER_N of Protomer 1 (chain B) could be modeled (Fig. 3). These MPER structures demonstrated strong similarity to those in Tri FPPR.2:

1. 1.

   The chain D MPER of Tri FPPR.1 adopts a structure similar to the helix-hinge-helix configuration of the MPER in chain B of Tri FPPR.2;

2. 2.

   Sharp turns between the α9 and MPER_N helices direct the latter towards the α9 helix of the adjacent protomer; the angle of this turn in the Tri FPPR.1 chain D is 74°, similar to that of the Tri FPPR.2 chain B;

3. 3.

   The hinge following the MPER_N helices positions MPER_C parallel to the α9 helix of the adjacent protomer and directs it towards the membrane;

4. 4.

   The Tri FPPR.1 chain D and Tri FPPR.2 chain B MPER_C segments enter the membrane in similar positions with respect to the Env ectodomain, opposite the gp41 tryptophan collar (residues 623-631) and α9 C-terminus (Glu 662-Asp 664) of the adjacent Env protomer;

5. 5.

   The gp120 C-terminus is interposed between the α9 helix of the same protomer and the MPER_N helix of the adjacent protomer;

6. 6.

   The MPER of Protomer 3 (chain F) is poorly ordered in both Tri FPPR.1 and Tri FPPR.2 structures.

In summary, although the MPERs of both Tri FPPR.1 and Tri FPPR.2 Envs are asymmetric, each conformer retains MPER chains with highly compatible structures.

The relationship of the Tri FPPR.1 and Tri FPPR.2 Envs to the A18 amphipol-lipid nanodisc was examined (Supplementary Fig. 8). Relative to the nanodisc axis, the Tri FPPR.1 and Tri FPPR.2 ectodomains are tilted 30 and 20 degrees, respectively. When the asymmetric Tri FPPR.1 and Tri FPPR.2 Env ectodomains are aligned, the direction of their tilts with respect to the nanodiscs differs by 155 degrees. As a result, different protomers in the Tri FPPR.1 and Tri FPPR.2 Envs are positioned closer to the nanodisc surface. Notably, the proximity of the Env protomer to the nanodisc is related to the degree of order in the associated MPER. This observation implies that MPER density is preserved best when the MPER is closer to the amphipol-lipid environment of the nanodisc.

The commonalities and potential complementarity of the Tri FPPR.1 and Tri FPPR.2 maps prompted us to use a combination of these maps to model a more complete MPER structure (Fig. 4a). The resulting composite Env model suggests that interlocking gp41 MPERs, gp41 α9 helices and the gp120 C-termini from all three protomers potentially form a membrane-proximal base that supports the rest of the Env ectodomain. Single-residue MPER changes that disrupt the Env PTC^{26,27,28,29,30} map to the MPER_N helix that, in the Tri FPPR structures, abuts the gp120 C-terminus and α9 helix from the adjacent protomer (Fig. 4b).

a To develop a more complete model of the MPER within the Env trimer structure, the 5 Å-filtered maps of Tri FPPR.1 and Tri FPPR.2 were aligned. The Tri FPPR.1 model was adjusted to fit the aligned maps, with helical fragments of the Chain B MPER-TM region from the Tri FPPR.2 model integrated accordingly, followed by refinement and modeling of the inter-helical loops to ensure structural coherence. In a similar way, the MPER_N helix of chain F was modeled into appropriately located, unexplained density apparent in the lowpass-filtered Tri FPPR.2 maps. The resulting composite model is shown from the perspective of the viral membrane. The Env subunits are colored according to the scheme in Fig. 1. The composite model highlights the interprotomer interactions involving the gp41 α9 and MPER helices and the gp120 C-termini (asterisks) that potentially stabilize the membrane-proximal base of Env. The locations of the last resolved gp120 C-terminal residues (Arg 511.A, Arg 511.C, Gln 507.E) are indicated by asterisks. b One face of the membrane-proximal base of the Tri FPPR.2 model is shown, fitted into the unfiltered density map. To enhance clarity, disconnected bits of density with a size smaller than 8.5 Å, as measured by the largest bounding box dimension (X, Y, or Z), are not shown. Changes in MPER residues that result in the disruption of the PTC³⁰ are highlighted in yellow on the ribbon diagram.

Env cleavage allows association of the gp120 C-terminus with the membrane-proximal base

The association of the gp120 C-terminus with the membrane-proximal base in the Tri FPPR Env structures suggested an explanation for the observation that proteolytic cleavage of the gp160 Env precursor stabilizes the PTC^22,23,24,25. To investigate the potential contribution of Env precursor cleavage to the formation of the hypothesized membrane-proximal base, we compared the structures surrounding the Env cleavage site in the uncleaved Env(-)²⁵ and cleaved AE2⁴¹ and Tri FPPR Env trimers (Fig. 5). As the surface-exposed cleavage site is disordered to various degrees in these Env structures, we measured the linear distance (LD) between the last resolved gp120 residue and the first resolved gp41 residue in each protomer of these Env trimers (Fig. 5a). LD represents a minimum estimate of the actual distance that must be traversed by the unresolved peptide encompassing the Env cleavage site, as steric barriers are not accounted for. If N represents the number of amino acid residues available to span the linear distance LD, LD/N represents the average span of an amino acid residue required to bridge the unresolved gap, in the absence of Env cleavage. LD/N cannot exceed the physical length limit for an amino acid residue (~3.4 Å) without Env cleavage having occurred. Both LD and LD/N were higher in the cleaved Envs than in the uncleaved Env(-) (Fig. 5a) (Mann-Whitney test, P = 0.02), indicating that Env cleavage results in the movement of one or both newly created termini away from each other. LD/N strongly correlated with the last resolved gp120 C-terminal residue (Fig. 5b, left panel), but not significantly with the first resolved gp41 N-terminal residue (Fig. 5b, right panel). This relationship is consistent with Env cleavage allowing the gp120 C-terminus to achieve a more ordered conformation. We note that the LD/N values varied among the cleaved Envs and even among the Tri FPPR protomers, observations that have several implications. First, the highest LD/N values far exceeded the amino acid physical limit, confirming that Env cleavage must have occurred to allow the observed gp120 C-terminal conformations in these Tri FPPR Env protomers. Second, the highest LD/N values were observed for the gp120 glycoproteins in chain C of Tri FPPR.1 and Chain A of Tri FPPR.2, whose C-termini potentially interact with the well-ordered MPER_N helices of their respective protomers (Fig. 5c). Tri FPPR Env protomers with lower LD/N ratios and less well-resolved gp120 C-termini have less ordered MPERs. Third, in the AE2 or sgp140 SOSIP.664 trimers, where the MPERs are completely disordered or absent, respectively, the gp120 C-termini were resolved only to Val 505, a residue that is positioned similarly in all cleaved Env trimer structures. These observations support a mechanism whereby Env cleavage frees the gp120 C-terminus, allowing the terminal six residues to interact with the cognate MPER, thereby stabilizing the membrane-proximal base and the pretriggered Env conformation (Fig. 5d).

a The linear distances (LDs) between the last resolved gp120 C-terminal residue and the first resolved gp41 N-terminal residue in the protomers of the uncleaved HIV-1_JR-FL Env(-) trimer (PDB ID: 7N6U)²⁵, the cleaved AE2 Env timer (PDB ID: 8FAE)⁴¹, the cleaved Tri FPPR.1 Env trimer and the cleaved Tri FPPR.2 Env trimer are shown. LD was measured from C_α atom to C_α atom of the indicated gp120 and gp41 amino acid residues. N is the number of residues available to fill the gap and equals the number of unresolved residues + 1. The linear distance per residue (LD/N) represents the average span of an amino acid residue required to bridge the unresolved gap surrounding the Env cleavage site, in the absence of cleavage. b The relationships between the last resolved gp120 C-terminal residue and LD/N (left panel) and between the first resolved gp41 N-terminal residue and LD/N (right panel) are shown. The distance spanned by an amino acid residue in a fully extended peptide strand (3.4 Å) is marked by the red horizontal line. For LD/N values that exceed this physical limit, Env cleavage would be required to obtain the conformations of gp120 and gp41 observed in the structures. The degree of order in the gp120 C-terminus, reflected in the extent of the residues resolved in the structures, strongly correlates with the LD/N values. The Spearman rank order correlation coefficients (r_S) and two-tailed P values are shown. c Views of the membrane-proximal region from the Tri FPPR.1 and Tri FPPR.2 Env structures are shown, illustrating the relationship of the ordered gp120 C-termini to the MPER_N helices of the gp41 subunits from the same protomer. In the Tri FPPR.1 model on the left, gp120.C (light green) and gp41.D (dark green) chains are shown. In the Tri FPPR.2 model on the right, gp120.A (salmon) and gp41.B (red) chains are shown. In both cases, based on LD/N values greater than 3.4 Å, the observed conformations of the gp120 C-termini in the membrane-proximal base of the Tri FPPR Env trimers could only occur after Env cleavage. d Env sequences and structural relationships near the gp120-gp41 cleavage site (black triangle) are summarized. Env sequences that are disordered in the analyzed structures are indicated by wavy lines. Note that the degree of order in the gp120 C ­terminus (residues 506-511) is dependent both on Env cleavage and on the degree of order in the MPER of the same protomer.

Structural correlates in cleaved, solubilized HIV-1 Env trimers

The HIV-1_AD8, AE2, Tri FPPR.1 and Tri FPPR.2 Env trimers were solubilized from cell membranes and prepared in different ways (Supplementary Fig. 1a), yet share a common asymmetric architecture. We compared these structures to identify generalizable relationships between specific structural parameters:

1. 1.

   Env tilting in the nanodisc and protomer asymmetry. In a previous study⁴¹, using the sole available AE2 Env structure, we noted that the Env ectodomain tilted in the SMA nanodiscs towards the interprotomer interfaces with the smaller opening angle. Examination of the Tri FPPR.1 and Tri FPPR.2 Envs indicates that this is not generalizable. Whereas the Tri FPPR.1 Env tilts towards the small opening angle, the Tri FPPR.2 Env tilts away from the small opening angle (Supplementary Fig. 8).

2. 2.

   MPER order/disorder and Env protomer asymmetry. In all four Env structures (AD8, AE2, Tri FPPR.1, and Tri FPPR.2), the map density for the gp41 MPER of chain F (Protomer 3) is weak, suggesting disorder. To achieve the observed asymmetric Env trimers from a C3-symmetric trimer, the shift required in Protomer 3, particularly in gp41/chain F, is greater than that in the other protomers (Supplementary Fig. 6a). As the MPERs of the other protomers are disordered in the AD8 and AE2 Envs, we were unable to evaluate a potential correlation between MPER order and protomer asymmetry. This correlation is apparent in the Tri FPPR Env structures. The MPER density is more ordered in chain B (Protomer 1) of Tri FPPR.1 and Tri FPPR.2, and in chain D (Protomer 2) of Tri FPPR.1 (Table 2). Thus, the MPER order is associated with protomers that require smaller shifts to achieve the observed asymmetry.

3. 3.

   Connectivity to an ordered MPER and the tilt angles of the α9 helices. In each protomer, the gp41 α9 helix is immediately N-terminal to the corresponding MPER. The tilt angles of the α9 helices relative to the gp120 subunit vary among the protomers of these asymmetric Env trimers. We used the α9 helix of chain F as a reference because it is not constrained by a connection with a well-ordered MPER in any of the studied Envs. We found that the tilt of the gp41 α9 helices is inversely related to the strength of the connection with the MPER. Where MPER density is strong, indicative of an ordered MPER, the tilt angles of the α9 helices are further from those of the chain F α9 helix (Supplementary Fig. 9, Table 2). Connectivity with a stable MPER apparently restrains the α9 helices from tilting.

4. 4.

   The tilt of the α9 helix and the formation of the HR1_N helix on the adjacent gp41 HR1_N region. Larger tilt angles of the α9 helices (with weaker MPER connectivity) correlate with the presence of helical HR1_N conformations on the clockwise adjacent protomer (viewed from the perspective of the viral membrane) (Supplementary Figs. 6b and 9, Table 2).

5. 5.

   Helicity of the gp41 HR1_N region and interprotomer opening angle. A strong direct correlation was observed between the helicity of the HR1_N region of a given gp41 protomer and the opening angle between that protomer and the counterclockwise adjacent protomer (viewed from the perspective of the viral membrane) (Supplementary Figs. 6b and 9b, Table 2). Helix-helix packing between HR1_N and α9 on the adjacent protomer should keep the interprotomer angle open.

These observations suggest a model in which Env tilting in the amphipathic copolymer-lipid nanodisc coincides with the disruption of the pretriggered state of one or two MPERs. A disordered MPER cannot maintain the conformation of the connected α9 helix, which is free to stabilize the HR1_N helix on the adjacent protomer. This α9-HR1_N interaction secures the opening of the interprotomer angle in the asymmetric Env trimer.

Comparison of the Tri FPPR MPER with other MPER structural models

We compared our Tri FPPR structures with available structural models of the MPER without MPER-targeted antibodies (Supplementary Table 1 and Supplementary Fig. 7a–c). The structure of an MPER peptide monomer in dodecyl phosphatidylcholine (DPC) micelles has been solved by NMR spectroscopy (PDB ID: 2PV6)⁵⁰. From the N- to the C-terminus, the MPER peptide consists of a two-turn alpha helix, a hinge, a one-turn helix and a 3₁₀ helix (Supplementary Fig. 7a, b). Conserved hydrophobic faces of these helical structures are buried in the DPC micelle, whereas more variable hydrophilic residues near the hinge and one-turn helix are solvent exposed. We aligned the 2PV6 MPER peptide with the chain B MPER from the Tri FPPR.2 Env (Supplementary Fig. 7b). Both structures comprise an amphipathic helical MPER_N followed by a hinge that directs the C-terminal MPER segment towards the membrane. However, the boundaries of the helices and the orientations of the MPER components differ between the structures. The 2PV6 N-terminal helix breaks earlier, at Trp 672, compared with the Tri FPPR MPER_N helix, which ends at Thr 676. The two structures diverge significantly around the hinge region, leading to different orientations of the C-terminal MPER elements.

By virtue of association with complete Env ectodomains or TM regions, a few unliganded MPER peptides form trimers that have been structurally characterized. Two NMR structures of MPER-TM peptides in nanodiscs or phospholipid bilayers^51,52 differ dramatically from one another; in 6E8W, the MPER-TM peptides assume a stalk-bubble conformation, whereas in 6DLN, the MPER-TM peptides assume a tripod conformation. Although both MPER-TM peptide structures contain helical elements, the helical boundaries and interhelical turns do not coincide with those of the Tri FPPR MPER_N and MPER_C helices, resulting in large differences in the trimeric MPER structures (Supplementary Fig. 7a, c). Likewise, in a cryo-EM structure (PDB ID: 7SC5) of a membrane SOSIP-modified Env reconstituted in nanodiscs⁴², the hypothetically modeled MPER_N turns from the α9 helix in a direction opposite to that seen in the Tri FPPR Envs, resulting in a dissimilar overall MPER structure. The thin stalks at the base of HIV-1 Env spikes visualized at lower resolution by cryo-electron tomography (cryo-ET)^53,54 contrast with the more substantial membrane-proximal base modeled for the Tri FPPR Env; we note that the aldrithiol (AT-2) used to inactivate the HIV-1 virions in these cryo-ET studies has been shown to contribute to destabilization of the State-1 Env conformation⁵³. In summary, none of the available structural models of unliganded trimeric MPERs fully align with the Tri FPPR MPERs, underscoring the highly context-dependent nature of MPER conformation.

Variation in the overall conformation of the Env base between the Tri FPPR Env and existing antibody-free MPER structures is reflected in interprotomer distances. For example, interprotomer Cα-Cα distances for Lys 665, near the N-terminal boundary of the MPER, are 24 Å for the two residues resolved in the Tri FPPR.1 model, compared with 53 Å in the 6DLN tripod, 16 Å in the 6E8W stalk-bubble and 30 Å in the 7SC5 membrane Env. Additional comparisons between the Tri FPPR composite model and available MPER peptide structures are shown in Supplementary Table 2. The Protomer 1-Protomer 2 distances in the composite model, which may be more typical of the native distances, differed significantly from those of the tripod-like 6DLN and stalk-bubble 6E8W MPER peptide structures. Although multiple variables potentially contribute to the observed differences, the connectivity of the MPER to the rest of the Env trimer ectodomain, particularly the direction of the turn arising from the α9 helix, likely influences the MPER conformation in the Tri FPPR Env.

Comparison of the Tri FPPR MPER structures with those of MPER peptides complexed with MPER bNAb fragments^55,56,57 revealed major differences in secondary and tertiary structure (See Supplementary Fig. 7a and d for examples). This is consistent with the MPER bNAbs recognizing primary HIV-1 Env conformations that are downstream of those observed in the Tri FPPR Env^17,18,56.

Glycan heterogeneity in the gp120 V1/V2 region

The densities associated with most of the glycans in the Tri FPPR.1 and Tri FPPR.2 maps were symmetrically distributed among the three Env protomers. However, 3D classification that was focused on one gp120 subunit from all protomers of the Tri FPPR Env maps yielded two major subclasses (designated gp120.1 and gp120.2) that exhibit heterogeneity in the gp120 V1/V2 region at the apex of the Env trimer. These subclasses are distributed equally between Tri FPPR.1 and Tri FPPR.2, suggesting that the variation of glycan structures is orthogonal to and uncoupled from the variation of MPER conformations. The gp120.1 map exhibits strong density associated with the V1 loop and with the Asn 130 and Asn 160 glycans, which stack against each other (Supplementary Fig. 10). The gp120.2 map exhibits stronger density associated with the glycans modifying Asn 156 and Asn 188. The representation of gp120.2 particles in the Tri FPPR Env population exceeded that of gp120.1 particles. Mixed Env trimers with at least two gp120.2 protomers were thus better represented, allowing the generation of higher-resolution maps (Supplementary Fig. 11). Permutations of the gp120.1/gp120.2 protomers, combined with the asymmetry of the Tri FPPR Envs, create substantial structural heterogeneity at the trimer apex. Such heterogeneity could diminish the generation or efficacy of V2 quaternary bNAbs, which bind the Env trimer apex in an asymmetric manner dependent on the glycans at Asn 156 and Asn 160^37,39,58,59.